We also applied a set of additional site filters as follows. We filtered out sites not on the MAC5 site list to restrict the site list to those that could be imputed well. We also filtered out sites if (i) any study (apart from 1000 Genomes) had a Hardy-Weinberg Equilibrium (HWE) p-value < 10-10, (ii) any study (apart from 1000 Genomes) had an overall inbreeding coefficient < -0.1, (iii) a MAF>0.1 with the site being called in fewer than 3 of the studies and not called in 1000 Genomes (the latter restriction kept sites present at high frequencies in non-European populations that were only called in 1000 Genomes). We also filtered out sites called only in the GoNLstudy or IBD cohort. We completely excluded GPC haplotypes from this step of the site list creation process.
